Cerebellar Structure and Function

Last modified: 2008-08-12 13:05:02 UTC

A very simple theory for describing the structure and function of the cerebellum is presented. This theory maintains that the cerebellum is functioning as a peak limiter, where its job is to smooth out spikey signals received from other structures in the brain.

Though the cerebellum only accounts for 10% of the volume of the brain, it contains an estimated 60% of the brain's neurons.¹ Obviously, something very important is going on within this structure.

It's also a brain structure that is present in all vertebrate species, including fish, reptiles, birds, and mammals.

Modern research into the structure and function of the cerebellum is finding that it is interconnected with most other parts of the brain, and as theories of the brain get more sophisticated, theories of the nature of processing occurring in the cerebellum are getting more complicated as well.

And yet we should know better. The structure of the cerebellum is extremely well-defined, and this same structure is present in the cerebella of all vertebrates. We should not be looking for a sophisticated way in which the human cerebellum might be interacting with human-specific cerebral structures. We should be looking for a generic utility that would be required by any brain.

This generic function is to attenuate the peaks in the output from cerebral structures.

The one problem that all vertebrate brains have in common is that cortical columns go into rapid bursting mode at unpredictable times, and where a coordinated muscular response is a sum effect of many cortical columns, the response tends to be very spikey.

Smoothing out spikey signals is a common problem in electronics, where the simplest method is to use peak limiter circuitry. Essentially, the signal is split into two pathways. Then one of the signals is attenuated by a fixed amount, and inverted, and then this signal is added back to the primary signal. The result is a steady output, regardless of fluctuations in the input.

This same effect can be accomplished easily with neurons in a specific configuration. In its simplest representation, all we have to do is branch off a secondary pathway, and then bring that pathway back in as inhibitory impulses. Here we want for the firing threshold of the secondary pathway to be higher than that of the primary pathway. If its threshold was the same as the other neurons in the sequence, then its inhibitory connections would cancel out the excitatory connections along the primary pathway, and there wouldn't be any output. But if the secondary pathway fires at a rate that is tied to the firing rate of the input, but is a fixed amount slower, the effect will be that of a peak limiter. (If the firing rate is proportionally lower instead of a fixed amount lower, then the affect is that of a dynamic range compressor, and that's of no use to us.)

The significance of all of this is that as we attempt to simulate more and more sophisticated neural nets, it will become important to make sure that the processing being done in the cerebellum is represented in the simulation. In consideration of the number of neurons in the cerebellum, if we have to allocate computing resources to cerebellar processing in exactly the same way that the brain does, then 60% of the clock cycles will be chewed up with that. But if the cerebellum is actually just functioning as a peak limiter, then we can accomplish this very easily in other ways. First, we could simply not bother to be quite so realistic in making the output from the primary motor cortex as spikey as the real thing. Second, if we're working on a robotics project, we could do the peak limiting as electronic (or even electric) post-processing, without taxing the CPUs that are doing the cognitive processing.